Listen to this article
Fashion retailers are increasingly turning to data analytics to keep up with the latest trends and client demands.
As well as having to meet the demands of “fast fashion” — customers wanting the latest designs from catwalk in stores the instant they appear — businesses must also price items correctly, know when to reduce them, stock enough of the right styles, colours, fabrics and sizes, and ensure that stores are well supplied and operate efficiently.
Data analytics is not new to the industry, which has long used spreadsheets and analysed sales information. However, new sources of data are now available, such as the information on mobile devices or social media sites. “The biggest change is the growth of unstructured data [data not stored in databases] — the texts, images, audio and YouTube videos,” says Keith Mercier, retail industry leader for global cognitive business solutions at IBM.
One method being deployed by retailers to discover more about what customers might want is the use of cognitive computing — programs that simulate human thought process and mimic the functions of the brain.
Cognitive computing relies on techniques such as data mining — the analysis of data from different sources — pattern recognition and natural language processing.
Mr Mercier says these types of applications mean vast new data sets can now be analysed, producing faster insights into fashion trends. “If we can give a retailer a two-week jump on trend prediction, [then] two weeks of selling time in stores is golden in this highly competitive industry,” he says.
By tracking how customers behave while shopping, data analytics can also help to improve the design and management of shops and department stores. Despite the growth of online fashion outlets, many consumers still visit stores to touch and try clothing or shoes before buying.
“The need [for companies] to know who is in the store — with [customer] permission — the moment they walk in is greater than it’s ever been,” says Brent Franson, chief executive of Euclid Analytics, a US-based company that uses location analytics to monitor consumer traffic in shops and malls.
Euclid uses WiFi signals from smartphones to track and analyse everything from the number of people entering a store to the length of time they stay and how often they come back. Customers can opt out of having data collected.
“Knowing your purchase history, and the kinds of things you buy, retailers can create a more personalised experience,” says Mr Franson.
Retailers also need to know when new products become available, which colours, styles or brands are selling fastest and when to mark down items.
These kinds of insights are what WGSN, a fashion trend forecaster, provides to clients through its Instock service, which tracks the clicks of online shoppers as they browse and buy items.
“It’s all about understanding when people are going into markdown, making sure you’re competitive on price and that you have the right balance of items,” says Francesca Muston, WGSN’s head of retail and product analysis.
Her company conducts catwalk analytics, with teams of experts tagging each outfit — noting its garment type, style, colour fabric and other details — as it is presented on the runway. Analysing this data reveals whether skirts or trousers are dominant in a particular season and, if it is trousers for example, whether they are mostly wide leg, flared-leg or bootleg styles.
Applying analytics to fashion is not easy, particularly as garments may have different names in different territories — trousers are pants in the US — and the lines between garment types are blurring with hybrids such as the “coatigan”, a softer or knitted version of a coat.
So while data analysis is a powerful tool, Ms Muston argues that it will never entirely replace human insights. “A lot of it comes down to intuition,” she says.
Nevertheless, the use of data is helping some businesses to adopt a more counterintuitive approach by designing algorithms that will choose people’s clothes for them. For example, no styles are shown on US-based online retailer Stitch Fix’s website. Instead, it sends shoppers a box of five items that have been selected according to the “style profile” users have created by answering questions on everything from their favourite colours and fabrics to their size, budget and lifestyle.
The San Francisco-based company employs 75 data scientists who have developed algorithms that aim to ensure that as few items as possible sent to customers will be returned.
After receiving their items, customers decide what to buy and what to return, and provide detailed feedback on the selection or what they would like to receive in future. “Those two sets of data — preference data and feedback data — drive everything,” says Eric Colson, Stitch Fix’s chief algorithms officer.
Despite such advances, Ms Muston believes many retailers need to do more to apply analytics to their own data.
“A lot of them are playing catch-up,” she says. “They need to understand how to better organise their systems so that their products can be classified and analysed and they can get a view on their data. That’s a big shift.”